The study, led by Assistant Professor Dowon Kim, PhD student Seongbeom Park, and Professor Jinhyeok Yun, shifts the focus from simple emission caps to atmospheric dispersion. While traditional methods rely on blanket speed reductions that often disrupt trade schedules, this new system uses physics-informed deep learning to reconstruct high-resolution flow fields. By analyzing how wind and weather transport pollutants, the framework identifies specific navigation windows that minimize public exposure.

Published in Ocean Engineering, the findings detail a process the team calls "temporal navigation." Instead of forcing uniform deceleration, the AI calculates precise speed and path adjustments based on current meteorological conditions. In simulations centered on Busan Port, this approach improved optimization performance by 20–35% compared to conventional strategies, while simultaneously cutting peak pollutant exposure in coastal communities by 34–78%. The technology offers a scalable path toward smarter port management and the integration of cleaner, autonomous maritime traffic systems.